Outlier calculator

Find outliers by the IQR method or by z-score, with every quartile, fence and score shown so the answer can be checked.

1 · Your numbers

Values

Separated by commas, spaces, tabs or new lines. Currency symbols are ignored; anything that is not a number is reported rather than dropped silently.

2 · Method

IQR multiplier

Z-score threshold

0Values

-Mean

-Median

-Std deviation

IQR method (Tukey's hinges)

Z-score method

Each value and its standard score
Value	Z-score	Flagged

Finding an outlier is not permission to delete it. A value outside the fences is a value worth looking at — it may be a typo, a unit mix-up or a failed sensor, or it may be the most interesting observation you have. Removing points because a formula flagged them is how real effects get erased. Decide with knowledge of where the data came from, and say what you removed and why.

Two methods, both shown in full

Paste your numbers and the tool gives you the quartiles, the interquartile range, Tukey's fences and every flagged value — plus the mean, standard deviation and z-score of each point. Nothing is a black box; every intermediate number is on screen so the result can be checked.

Which quartile convention

This matters, and most tools do not say. There is no single correct definition of a quartile, and different software gives different answers for the same data. This uses Tukey's hinges: sort the values, split at the median, and if the count is odd the median goes in neither half. That is the definition the box plot and the 1.5 x IQR rule were built on.

Excel's QUARTILE.INC and R's default interpolate instead and will give slightly different numbers on small samples. Neither is wrong; they answer a slightly different question. Knowing which one you are looking at is what lets you explain a difference instead of arguing about it.

The z-score method has a trap

An extreme value inflates the very mean and standard deviation it is being measured against, so a single large outlier can hide itself. Worse, on a small sample the score is mathematically bounded: with ten values no z-score can exceed about 2.85, so the usual threshold of 3 can never flag anything however wrong a value looks. The tool detects that case and says so rather than reporting a clean dataset.

Finding an outlier is not permission to delete it

This is the most important thing on the page. A value outside the fences is a value worth looking at. It might be a typo, a unit mix-up, or a sensor that failed — or it might be the single most interesting observation you have. Removing points because a formula flagged them is how real effects get erased. Decide with knowledge of where the data came from, and say what you removed and why.